CN105954709A - Acoustic vector circular array source number detection method based on characteristic value multiple threshold correction - Google Patents

Abstract

The invention belongs to the acoustic vector sensor array signal processing field, and more specifically relates to an acoustic vector circular array source number detection method based on characteristic value multiple threshold correction which is applied to underwater target remote passive detection. The method comprises establishing an acoustic vector circular array signal receiving model, obtaining acoustic vector circular array receiving acoustic pressure data, radial vibration velocity data and tangential vibration velocity, constructing a covariance matrix of acoustic vector circular array acoustic pressure and vibration velocity combined treatment, and decomposing characteristic values; and performing multiple threshold division treatment on a characteristic value set obtained by decomposing the covariance matrix to obtain a signal and noise corresponding characteristic value set. The method dynamically integrates an information theory detection method based on characteristic value multiple thresholds and the sound anti-noise performance of an acoustic vector circular array, obviously reduces the signal to noise ratio threshold of a detection algorithm, and overcomes the disadvantages that traditional detection methods such as MDL, diagonal loading MDL, and GDE are more sensitive to noise characteristic value change.

Description

A kind of acoustic vector circle battle array information source number detection method of feature based value multi thresholds correction

Technical field

The invention belongs to acoustic vector-sensor array column signal process field, be specifically related to relate to the acoustic vector circle battle array information source number detection method of the feature based value multi thresholds correction of a kind of long-range passive detection being applied to submarine target.

Background technology

Sources number estimation problem is a major issue in Array Signal Processing, high-resolution Estimation of Spatial Spectrum technology is it is generally required to first accurately estimate information source number, otherwise will cause orientation algorithm for estimating hydraulic performance decline, therefore have a wide range of applications in fields such as radar, sonar, communications.

Along with the development of acoustic vector sensors technology, vector hydrophone is widely used in each field of Underwater Acoustics Engineering.Vector hydrophone arrays signal processing can be effectively improved submarine target remote probe ability, and in recent years, the Sources number estimation problem of acoustic vector array also receives the concern of people.Li Nan et al. proposes a kind of acoustic vector sensor array Sources number estimation method (Li Nan unanimously differentiated based on core, Cheng Jin room, He Guangjin, Zhang Wei. Estimation Methods for Source Number [J] based on vector array. Wuhan University of Technology's journal, 2013,37 (1): 175-178.), the method utilizes parallel factor to analyze method representation signal covariance tensor, determines information source number by the order calculating this model；Zhang Kai propose a kind of based on second order blind source separation algorithm and resolve the acoustic vector sensor array number of source of vibration velocity model estimate and orientation algorithm for estimating (the iron of fine quality. number of source based on acoustic vector sensor array detection and orientation algorithm for estimating [J]. modern navigation, 2015,3:269-275.), similarity degree between acoustic pressure and parsing vibration velocity waveform that this algorithm obtains according to blind source separating, it is achieved number of source is estimated；But said method also underuses the anti-noise ability that in acoustic vector sensor array, acoustic pressure processes with vibration velocity united information, and the power of test of low signal-to-noise ratio is limited, also cannot meet submarine target long-range passive detection needs.

The substantially not difference of vector hydrophone arrays signal processing and traditional pressure hydrophone Array Signal Processing, but its key technology is how to make full use of the anti-noise ability that acoustic pressure processes with vibration velocity united information.Bai Xingyu et al. proposes the detection of acoustic vector sensor array information source number and the Subspace partition criterion (Bai Xingyu of a kind of feature based vector, Jiang Yu, Zhao Chunhui. acoustic vector sensor array information source number based on acoustic pressure vibration velocity Combined Treatment detection estimates [J] with orientation. acoustic journal, 2008,33 (1): 56-61.), the anti-noise ability of the high resolution of subspace method with acoustic vector sensor array is combined, reduce accessible signal-noise ratio threshold, it is achieved that the information source number detection of remote object is estimated with orientation；Zhang Kai propose number of source based on acoustic pressure, the acoustic vector sensor array canonical correlation technique of vibration velocity Combined Treatment estimate (the iron of fine quality. number of source based on acoustic vector sensor array canonical correlation technique estimates [J]. marine electronic resists, 2013,36 (2): 69-77.), canonical correlation technique is processed combine with acoustic pressure, vibration velocity information consolidation by this algorithm, it is achieved that signal number purpose effectively detects.But said method is only applicable to acoustic vector line array Sources number estimation, due to acoustic vector circle battle array signal subspace and noise subspace eigenvalue characteristic distributions, said method cannot directly apply to acoustic vector circle battle array.Uniform circular array have 360 ° comprehensive, estimate performance without fuzzy target detection and orientation, tool has an enormous advantage, and therefore has to compare in sonar and radar system and is widely applied.Compared with the achievement in research of acoustic vector linear array, acoustic vector circle battle array information source number context of detection achievement in research has no report.The present invention proposes the acoustic vector circle battle array information source number detection method of a kind of feature based value multi thresholds correction, the method overcome tradition Minimum description length criterion (MDL), diagonal angle loads the detection methods such as MDL, Gai Shi loop truss criterion (GDE) and changes noise characteristic value more sensitive, and the shortcoming that acoustic vector circle battle array acoustic pressure vibration velocity combination treatment method cannot be applied.

Summary of the invention

It is an object of the invention to the information source number detection performance proposing to improve under Low SNR, the acoustic vector circle battle array information source number detection method of the feature based value multi thresholds correction of the long-range passive detection of submarine target can be realized.

The object of the present invention is achieved like this:

The acoustic vector circle battle array information source number detection method of feature based value multi thresholds correction, on the basis of acoustic vector circle battle array acoustic pressure vibration velocity Combined Treatment, the division to covariance matrix eigenvalue and correction is realized by arranging different threshold values, and realize information source number detection by theory of information detection criteria, comprise the steps:

(1) acoustic vector circle battle array Signal reception model is set up, it is thus achieved that acoustic vector circle battle array receives acoustic pressure data P (t), radially vibration velocity data V_r(t), tangential vibration velocityThe covariance matrix R of structure acoustic vector circle battle array acoustic pressure vibration velocity Combined Treatment_pv, to R_pvCarry out Eigenvalues Decomposition；

(2) to covariance matrix R_pvCharacteristic value collection C obtained after decomposition_λCarry out multi thresholds division process, it is thus achieved that signal and noise character pair value set C_s、C_n；

(3) to noise character pair value set C_nAverage correcting process, obtain eigenvalue λ after noise correspondence correction_n；

(4) theory of information detection criteria MDL is used to realize information source number detection.

Described step (1) includes the radially and tangentially component of acoustic vector circle a burst of unit vibration velocity is received data V_r(t)、Project on the x in cartesian coordinate system xOy plane, y-axis direction, obtain acoustic vector circle vibration velocity x of battle array, y channel signal V_x(t)、V_y(t), and by rotary electronic obtain combine vibration velocity:

In formula, φ, for specifying observed direction, can use Givens conversion to determine；

According to acoustic pressure and vibration velocity Combined Treatment average acoustic energy stream concept, it is thus achieved that acoustic vector circle battle array Cross-covariance:

R_pv=E [P (t) V_c ^H(t)]

To covariance matrix R_pvCarry out Eigenvalues Decomposition:

By eigenvalue λ_iAscending arrangement forms characteristic value collection C_λ。

Described step (2) includes, to covariance matrix R_pvCharacteristic value collection C obtained after decomposition_λCarry out multi thresholds division process, it is thus achieved that signal and noise character pair value set C_s、C_n；Secondly to noise characteristic of correspondence value set C_nAverage correcting process, obtain eigenvalue λ after noise correspondence correction_n；Then use theory of information detection criteria MDL to realize information source number detection, specifically comprise the following steps that

(2.1) two threshold values L varied in size are set_t1、L_t2, to characteristic value collection C_λDivide, obtain signal characteristic value set C_s1(λ_i≥L_t1), the disputed areas C_Dis(L_t1＞ λ_i≥L_t2) and noise characteristic value set C_n2(L_t2＞ λ_i)；

(2.2) characteristic value collection C to be revised is defined_Rev=C_Dis∪C_n2, by C_RevThe ascending arrangement of middle eigenvalue；One new threshold value L is set_tiIf, threshold value L_ti＞ L_t2, then threshold value L can be passed through_tiTreat revisory eigenvalue set C_RevEigenvalue repartition, form two characteristic value collection C_siAnd C_ni, then characteristic value collection C can be obtained_λThrough the revised signal of multi thresholds and noise character pair value set C_s=C_s1∪C_si、C_Rev=C_ni；If threshold value L_ti≤L_t2, then revise and terminate；

(2.3) if threshold value L_ti＞ L_t2, repeat step (2) and repeatedly divide, form signal character pair value set C_sWith noise character pair value set C_n=C_Rev；

(2.4) to noise characteristic of correspondence value set C_nAverage correcting process, seek characteristic value collection C_nIn the meansigma methods of eigenvalueTo meansigma methodsCarry out loading processing, obtain eigenvalue after noise correspondence correction

(2.5) theory of information detection criteria MDL is used to realize information source number detection.

The beneficial effects of the present invention is: the theory of information detection method of feature based value multi thresholds correction is combined by the method with the acoustic vector circle good noiseproof feature of battle array, significantly reduce the signal-noise ratio threshold of detection algorithm, overcome traditional MDL, diagonal angle loads the shortcoming that the detection methods such as MDL, GDE are more sensitive to noise characteristic value change；It addition, the present invention still has good detection performance when the power of each information source of multi-target detection and array received exists certain difference.

Accompanying drawing explanation

Fig. 1 acoustic vector circle battle array information source number detection method flow chart；

Fig. 2 acoustic vector circle battle array lays schematic diagram；

Fig. 3 acoustic pressure, the vector processing method impact analysis result to detection performance；

The simulation analysis result of Fig. 4 algorithms of different detection performance；

The simulation analysis result of Fig. 5 multi-target detection performance；

Fig. 6 acoustic pressure, the analysis of experiments result of vector processing method；

The analysis of experiments result of Fig. 7 algorithms of different detection performance.

Detailed description of the invention

Below in conjunction with the accompanying drawings the present invention is described further.

The present invention relates to the acoustic vector circle battle array information source number detection method of a kind of feature based value multi thresholds correction, the method constructs acoustic vector circle battle array acoustic pressure and the covariance matrix of vibration velocity Combined Treatment, covariance matrix is carried out Eigenvalues Decomposition and obtains characteristic value collection；Based on acoustic vector circle battle array signal subspace and noise subspace eigenvalue characteristic distributions, characteristic value collection is divided by the threshold value varied in size by setting, obtains signal characteristic value set and noise characteristic value set；Noise characteristic value set is averaged correcting process, noise characteristic value after being revised；Then theory of information detection criteria MDL is used to realize information source number detection.The theory of information detection method of feature based value multi thresholds correction is combined by the method with the acoustic vector circle good noiseproof feature of battle array, significantly reduce the signal-noise ratio threshold of detection algorithm, and overcome traditional MDL, diagonal angle loads the shortcoming that the detection methods such as MDL, GDE are more sensitive to noise characteristic value change.Theoretical simulation and result of the test show that the present invention has more preferable noise inhibiting ability and target detection performance, and remote object passive detection aspect has good superiority under water.

First the method constructs the covariance matrix of acoustic vector circle battle array acoustic pressure and vibration velocity Combined Treatment, and the eigenvalue of covariance matrix is carried out multi thresholds correcting process, then utilizes theory of information detection criteria to achieve the information source number detection of acoustic vector circle battle array.

The technical solution adopted for the present invention to solve the technical problems comprises the following steps:

(1) acoustic vector circle battle array Signal reception model is set up, it is thus achieved that acoustic vector circle battle array receives acoustic pressure data P (t), radially vibration velocity data V_r(t), tangential vibration velocityThe covariance matrix R of structure acoustic vector circle battle array acoustic pressure vibration velocity Combined Treatment_pv, to R_pvCarry out Eigenvalues Decomposition；

(2) to covariance matrix R_pvCharacteristic value collection C obtained after decomposition_λCarry out multi thresholds division process, it is thus achieved that signal and noise character pair value set C_s、C_n；

(3) to noise characteristic of correspondence value set C_nAverage correcting process, obtain eigenvalue λ after noise correspondence correction_n；

(4) theory of information detection criteria MDL is used to realize information source number detection.

The present invention is further described with example below in conjunction with the accompanying drawings, and information source number detection method flow chart of the present invention is as it is shown in figure 1, specific embodiments is as follows:

The first step, sets up acoustic vector circle battle array Signal reception model, it is thus achieved that acoustic vector circle battle array receives acoustic pressure data P (t), radially vibration velocity data V_r(t), tangential vibration velocityThe covariance matrix R of structure acoustic vector circle battle array acoustic pressure vibration velocity Combined Treatment_pv, to R_pvCarry out Eigenvalues Decomposition.Comprise the following steps that described:

(1) assuming that M unit acoustic vector circle battle array is positioned at xOy plane, radius is r, using x direction as 0 ° of direction of circle battle array.The acoustic vector sensors vibration velocity component positive direction of principal axis of x, y passage lays, as shown in Figure 2 respectively along the circle radial direction of battle array, tangential direction.For shallow sea remote target acquisition, the most do not consider vibration velocity vertical component, only consider vibration velocity horizontal component.Assume have K incoherent arrowband sound-source signal S, far field (t) to incide in acoustic vector circle battle array, then acoustic vector circle battle array reception data:

In formula, P (t), V_r(t)、It is respectively sound pressure signal, radially vibration velocity signal and tangential vibration velocity signal that acoustic vector circle battle array receives；N_p(t)、N_vr(t)、It is respectively the acoustic pressure of isotropic noise, radially vibration velocity and tangential vibration velocity component that acoustic vector circle battle array receives.They can be expressed as:

S (t)=[s₁(t),…,s_K(t)]^T

In formula:

In formula, a_pmRepresent the steering vector of the sound pressure signal of m-th acoustic vector sensors reception, a_vrm、Represent the steering vector that m-th acoustic vector sensors vibration velocity radial direction, tangential component are corresponding respectively；θ_iRepresent i-th sound-source signal incident angle (i=1,2 ..., K), λ represents signal wavelength.

(2) by acoustic vector circle a burst of unit vibration velocity tangentially and radially component reception data projection to x, y-axis direction in cartesian coordinate system xOy plane, then m-th array element reception vibration velocity signal becomes:

In formula, v_rm、Represent that m-th acoustic vector sensors vibration velocity radial direction, tangential component receive signal respectively.After projective transformation, acoustic vector circle battle array receives data and is transformed to:

In formula, V_x(t)、V_yT () represents acoustic vector circle vibration velocity x of battle array, y channel signal after projection respectively.

Can obtain combining vibration velocity V by rotary electronic_c(t)、V_s(t):

In formula, φ, for specifying observed direction, can use Givens conversion to determine.

(3) according to average acoustic energy stream concept, the Cross-covariance of acoustic pressure vibration velocity Combined Treatment can be obtained:

R_pv=E [P (t) V_c ^H(t)] (4)

To covariance matrix R_pvCarrying out Eigenvalues Decomposition is following form:

In formula, E [] expression takes statistical average, λ_iAnd u_iFor ith feature value and the vector of correspondence, and by eigenvalue λ_iAscending arrangement forms characteristic value collection C_λ。

Second step, to covariance matrix R_pvCharacteristic value collection C obtained after decomposition_λCarry out multi thresholds division process, it is thus achieved that signal and noise character pair value set C_s、C_n.Specifically comprise the following steps that

(1) two threshold values L varied in size are set_t1、L_t2, to characteristic value collection C_λDivide, obtain signal characteristic value set C_s1(λ_i≥L_t1), the disputed areas C_Dis(L_t1＞ λ_i≥L_t2) and noise characteristic value set C_n2(L_t2＞ λ_i)；Threshold value L_t1、L_t2It is respectively as follows:

(2) characteristic value collection C to be revised is defined_Rev=C_Dis∪C_n2。

(3) by C_RevThe ascending arrangement of middle eigenvalue；One new threshold value L is set_tiFor:

If (i) threshold value L_ti＞ L_t2, then threshold value L can be passed through_tiTreat revisory eigenvalue set C_RevEigenvalue repartition, form two characteristic value collection C_siAnd C_ni, then characteristic value collection C can be obtained_λThrough the revised signal of multi thresholds and noise character pair value set C_s=C_s1∪C_si、C_Rev=C_ni；

(ii) if threshold value L_ti≤L_t2, then revise and terminate.

(4) if threshold value L_ti＞ L_t2, repeat step (3) and repeatedly divide, form signal character pair value set C_sWith noise character pair value set C_n=C_Rev。

3rd step, to noise characteristic of correspondence value set C_nAverage correcting process, obtain eigenvalue λ after noise correspondence correction_n.Specifically comprise the following steps that

(1) characteristic value collection C is sought_nIn the meansigma methods of eigenvalue

(2) to meansigma methodsCarry out loading processing, noise character pair value after being revisedAccording to theoretical simulation and experimental analysis, it is thus achieved that loaded value is:

Add=fix (4 min{ λ_i|λ_i∈C_n}) (8)

In formula, fix () represents downward round numbers.

4th step, uses theory of information detection criteria MDL to realize information source number detection.

MDL detection criteria computing formula is as follows:

In formula, L is hits, and n is signal number (degree of freedom) to be estimated, and Λ (n) likelihood function is:

Above the detailed description of the invention of summary of the invention each several part is illustrated.It is embodied as being described further to the present invention below by simulation example and test examples.

Simulation example:

Assume that acoustic vector uniform circular array unit number is 12 yuan, battle array radius r=0.7 λ, information source incident direction 60 °, fast umber of beats 1k, it is intended that observed direction 60 °.Detecting factor D (L) value of lid formula loop truss criterion (GDE) is 1.

Fig. 3 is acoustic pressure, the vector processing method impact analysis result to detection performance, and acoustic pressure vibration velocity Combined Treatment noiseproof feature is better than acoustic pressure and vector separate processing approach.Fig. 4 is the simulation analysis result of algorithms of different detection performance, and the inventive method detects performance under low signal-to-noise ratio and is better than additive method.Fig. 5 is the simulation analysis result of multi-target detection performance, and the inventive method detects the MDL method that performance is better than loading based on diagonal angle under this condition.

Test examples:

Completing the test of undersea long target information source number test experience at anechoic tank, experiment acoustic vector circle battle array is 8 yuan of acoustic vectors circle battle array (as shown in Figure 6), and battle array radius is 0.35m.Each acoustic vector sensors x and y positive direction respectively with the radial direction at this, tangentially overlap.In experiment, target sound source launches simple signal, and sound source, away from acoustic vector circle battle array about 16m, is positioned at 6 bugle call vector sensor v_xDirection.Filter passband frequency 500～5000Hz.Definition signal to noise ratio snr is:

In formula,WithIt is respectively sound source and pond Background Noise Power.It is 1k that data process fast umber of beats, it is intended that observed direction is 220 °.

Fig. 7 is the analysis of experiments result of acoustic pressure, vector processing method.

The analysis result of simulation example and test examples shows: (1) present invention utilizes and have employed P-V covariance matrix building method based on acoustic vector circle battle array acoustic pressure vibration velocity Combined Treatment, more traditional acoustic pressure processing method and acoustic vector sensors vibration velocity channel has the ability of more preferable anti-isotropic noise as the processing method of independent array element；(2) present invention uses theory of information detection method and acoustic vector circle battle array acoustic pressure, the good noiseproof feature of vibration velocity Combined Treatment of eigenvalue multi thresholds correction, significantly enhancing the low signal-to-noise ratio power of test of acoustic vector circle battle array information source number, remote object passive detection aspect has good superiority under water.

Claims (3)

1. the acoustic vector circle battle array information source number detection method of a feature based value multi thresholds correction, it is characterised in that: at acoustic vector circle On the basis of battle array acoustic pressure vibration velocity Combined Treatment, realize the division to covariance matrix eigenvalue and correction by arranging different threshold values, And realize information source number detection by theory of information detection criteria, comprise the steps:

(1) acoustic vector circle battle array Signal reception model is set up, it is thus achieved that acoustic vector circle battle array receives acoustic pressure data P (t), radially vibration velocity data V_r(t), tangential vibration velocityThe covariance matrix R of structure acoustic vector circle battle array acoustic pressure vibration velocity Combined Treatment_pv, to R_pvCarry out Eigenvalues Decomposition；

(2) to covariance matrix R_pvCharacteristic value collection C obtained after decomposition_λCarry out multi thresholds division process, it is thus achieved that signal and making an uproar Sound character pair value set C_s、C_n；

(3) to noise character pair value set C_nAverage correcting process, obtain eigenvalue λ after noise correspondence correction_n；

(4) theory of information detection criteria MDL is used to realize information source number detection.

The acoustic vector circle battle array information source number detection method of a kind of feature based value multi thresholds correction the most according to claim 1, its It is characterised by: described step (1) includes the radially and tangentially component of acoustic vector circle a burst of unit vibration velocity is received data V_r(t)、Project on the x in cartesian coordinate system xOy plane, y-axis direction, obtain acoustic vector circle vibration velocity x of battle array, y passage letter Number V_x(t)、V_y(t), and by rotary electronic obtain combine vibration velocity:

$\left\{\begin{array}{c}{V}_c\left(t\right)=V_x\left(t\right)cos\mathrm{\φ}+V_y\left(t\right)sin\mathrm{\φ}\\ V_s\left(t\right)=-V_x\left(t\right)sin\mathrm{\φ}+V_y\left(t\right)\cos \mathrm{\φ}\end{array}\right.$

In formula, φ, for specifying observed direction, can use Givens conversion to determine；

According to acoustic pressure and vibration velocity Combined Treatment average acoustic energy stream concept, it is thus achieved that acoustic vector circle battle array Cross-covariance:

R_pv=E [P (t) V_c ^H(t)]

To covariance matrix R_pvCarry out Eigenvalues Decomposition:

$R_{pv}=E\[P\left(t\right)V_c^H\left(t\right)\]=\underset{i=1}{\overset{M}{\Σ}}{\mathrm{\λ}}_i{u}_i{u}_i^H=\underset{i=1}{\overset{K}{\Σ}}{\mathrm{\λ}}_i{u}_i{\mathrm{\μ}}_i^H$

By eigenvalue λ_iAscending arrangement forms characteristic value collection C_λ。

The acoustic vector circle battle array information source number detection method of a kind of feature based value multi thresholds correction the most according to claim 1, its It is characterised by: described step (2) includes, to covariance matrix R_pvCharacteristic value collection C obtained after decomposition_λCarry out multi thresholds to draw Divisional processing, it is thus achieved that signal and noise character pair value set C_s、C_n；Secondly to noise characteristic of correspondence value set C_nPut down Homogenizing correcting process, obtains eigenvalue λ after noise correspondence correction_n；Then theory of information detection criteria MDL is used to realize information source number Detection, specifically comprises the following steps that

(2.1) two threshold values L varied in size are set_t1、L_t2, to characteristic value collection C_λDivide, obtain signal characteristic value collection Close C_s1(λ_i≥L_t1), the disputed areas C_Dis(L_t1＞ λ_i≥L_t2) and noise characteristic value set C_n2(L_t2＞ λ_i)；

(2.2) characteristic value collection C to be revised is defined_Rev=C_Dis∪C_n2, by C_RevThe ascending arrangement of middle eigenvalue；Arrange one New threshold value L_tiIf, threshold value L_ti＞ L_t2, then threshold value L can be passed through_tiTreat revisory eigenvalue set C_RevEigenvalue again draw Point, form two characteristic value collection C_siAnd C_ni, then characteristic value collection C can be obtained_λThrough the revised signal of multi thresholds and noise pair Answer characteristic value collection C_s=C_s1∪C_si、C_Rev=C_ni；If threshold value L_ti≤L_t2, then revise and terminate；

(2.3) if threshold value L_ti＞ L_t2, repeat step (2) and repeatedly divide, form signal character pair value set C_sWith make an uproar Sound character pair value set C_n=C_Rev；

(2.4) to noise characteristic of correspondence value set C_nAverage correcting process, seek characteristic value collection C_nIn eigenvalue Meansigma methodsTo meansigma methodsCarry out loading processing, obtain eigenvalue after noise correspondence correction

(2.5) theory of information detection criteria MDL is used to realize information source number detection.